It seems like they haven't been overzealous and cross-referenced hits with other data. People are getting varying degrees of bans, and exploiters with several accounts are reporting that not every account has been banned. i.e. Valve are only banning when they are certain.
I imagine they are looking at the honeypot, and in-game actions that would be a result of the player having information they shouldn't.
Unlikely that they checked each of the 40,000 bans individually, but I imagine they devised a simple quantitative check that they could automate like "honeypot = true, check how far from STDDev player's dewarding accuracy was", then they spotchecked the highest confidence rates until they were happy to rollout the banwave.
I imagine they are looking at the honeypot, and in-game actions that would be a result of the player having information they shouldn't.
Unlikely that they checked each of the 40,000 bans individually, but I imagine they devised a simple quantitative check that they could automate like "honeypot = true, check how far from STDDev player's dewarding accuracy was", then they spotchecked the highest confidence rates until they were happy to rollout the banwave.